okay... so i work in two way satellite and will expand a bit on what this looks like in real world practice, with currently available technology and satellite architectures.

geostationary orbit dedicated (1:1) transponder capacity, in real world use with large earth stations and SCPC modems, has latency of around 485ms to 490ms for a return trip ping. That's a figure from remote earth station (example: Nauru) to the other end of a link (example: Singapore).

119 x 4 = 476 ms

add anywhere from 10ms to 20ms for a combination of modem Tx encoding, framing, adding FEC, and then the reverse on the Rx side.

example:

nauru router with a 1000BaseLX link into a satellite modem generates one ICMP ping

nauru --> satellite -->

satellite --> singapore

singapore modem does decode, passes IP traffic to a router or something else that can answer ICMP

singapore --> satellite

satellite --> nauru

nauru satellite modem decodes, passes traffic to router

I was very careful not to draw any conclusion besides "closer" because this is not my wheelhouse.

Some of the low cost wireless broadband projects have proposed the use of a ground station fairly nearby as the upstream network provider but I haven't the fuzziest what SpaceX has planned.

In your opinion what fraction of the overall delays in the network are attributable to the diameter of the orbit (great circle distance and ground-to-satellite distance) instead of the limitations of the hardware?

delays in geostationary orbit based systems are limited by basically two things... one is the distance, assuming dedicated 1:1 capacity. From a consumer internet user perspective the other delays are oversubscription and TDMA timeslice related. Consumer grade VSAT-based internet services (hughnesnet, viasat, etc) that you can buy for a cabin in a far rural part of Idaho are oversubscribed at a 16:1, 32:1 or greater ratio. So the RTT latency will be anywhere from 495ms in the middle of the night up to 1250ms during peak evening hours. This is a network architecture limitation/capacity limitation at layers 2 and 3, not so much about the physical distance that the RF has to travel.

The SpaceX system will likely operate somewhat similarly to o3b, which has a number of backbone link earth stations around the earth, adjacent to major fiber IX points. They don't want to endlessly pass data satellite-to-satellite-to-satellite because that will reduce the overall throughput and capacity of the system. Within a single moving spot beam I bet they are designing the system to pass data satellite-to-satellite if necessary. Or as few hops as possible. Once again using a rural Idaho example, at any given time a satellite or an adjacent pair of satellites in two orbital planes might "see" both a rural cabin CPE, and also a larger earth station/trunk link located next to terrestrial fiber in Boise. The goal is to get the traffic from the CPE to Boise in a few hops as possible. Ideally in a single "bent pipe" relay architecture.

I’m still not sure if you answered my question but thank you for participating. We don’t often get people with real experience in satcom.

I’ve never had to understand satellite physics beyond explaining to a director why he needed terrestrial internet to solve his telecommunications problem. They get the Grace Hopper explanation (aka the speed of light says data can only move this far per nanosecond)

Space X is actually aiming for a low earth orbit constellation. I've heard others mention numbers of around 50ms which is quite acceptable in my mind.

But how does that 50ms factor into the overall latency experienced by the end user?

As I recall, SpaceX will route user<->sat1<->sat2<->server. At low orbit, rtt for sat1<->sat2 is comparable to that for user<->server. Maybe even less, because ~vacuum vs fiber or copper. So you just add 50 msec to the landline rtt.